January 2007 - usocket-devel - mailman3.common-lisp.net

[usocket-devel] usocket 0.2.3 bug (openmcl 1.0/darwin32 backend)
by kevin montuori 13 Jan '07

13 Jan '07

hi -- first, thanks for a very useful library. i've run across a bug in the usocket 0.2.3 package, specifically in backend/openmcl.lisp. the openmcl-socket:socket-connect function takes exactly one argument, socket.[1] the usocket code provides two arguments, socket and :element-type. it appears (networking is not necessarily my strong point!) that with sockets of type :stream (i.e., not :datagram) the format is always bivalent, making the :element-type distinction inconsequential.[2] i've provided diff -c output[3] for the fix that i required. please excuse me if this has already been reported (or fixed) already, i didn't see mention of it. cheers. k. [1] http://openmcl.clozure.com/Doc/index.html#SOCKET_002dCONNECT [2] http://openmcl.clozure.com/Doc/index.html#Overview-_003c3_003e [3] *** openmcl.lisp.orig Sat Jan 13 09:18:21 2007 --- openmcl.lisp Sat Jan 13 09:25:42 2007 *************** *** 45,54 **** (let ((mcl-sock (openmcl-socket:make-socket :remote-host (host-to-hostname host) :remote-port port))) ! (openmcl-socket:socket-connect mcl-sock ! :element-type (if (subtypep element-type ! 'character) ! :text :binary)) (make-stream-socket :stream mcl-sock :socket mcl-sock)))) (defmethod socket-close ((usocket usocket)) --- 45,51 ---- (let ((mcl-sock (openmcl-socket:make-socket :remote-host (host-to-hostname host) :remote-port port))) ! (openmcl-socket:socket-connect mcl-sock) (make-stream-socket :stream mcl-sock :socket mcl-sock)))) (defmethod socket-close ((usocket usocket)) -- kevin montuori montuori(a)gmail.com this signature no verb.

2 1

[usocket-devel] Why rely on implementation specific network code?
by Hedos 05 Jan '07

05 Jan '07

Why not make a CFFI library based around the C sockets library? Wouldn't that be easier? That would avoid running into some implementation limitations too.

2 2

[usocket-devel] [ANNOUNCE] usocket 0.2.2
by Erik Huelsmann 03 Jan '07

03 Jan '07

Just a minute ago, I released usocket version 0.2.2, which: - Adds ECL support - Fixes an SBCL bug With now 9 Common Lisp implementations supported, I hope usocket proves to be making good on its promise of being a portability layer. If you use or plan to use usocket in your library or application, please send feedback to usocket-devel(a)common-lisp.net! Erik Huelsmann

1 0

[usocket-devel] GCL socket code on slime-devel
by Erik Huelsmann 02 Jan '07

02 Jan '07

Hi! I found a mail from you on slime-devel where you have started a slime backend which includes some socket code for GCL. I'd like to extend usocket (http://common-lisp.net/project/usocket/) to support GCL and you have already done 90% of that work, so, I'd like to have your permission and take the code (if it is yours to give) into usocket. What's the license you put on it? usocket is currently under MIT license. Does your contribution as linked to in the link below fall into the same category (public domain) as the rest of SLIME? http://common-lisp.net/pipermail/slime-devel/2005-April/003392.html Thanks in advance! bye, Erik.

1 0